Improving nonstoquastic quantum annealing with spin-reversal transformations

Nonstoquastic Hamiltonians are hard to simulate due the sign problem in quantum Monte Carlo simulation. It is, however, unclear whether nonstoquasticity can lead advantage annealing. Here we show that $YY$ interactions between qubits make adiabatic path during annealing, and therefore performance, dependent on spin-reversal transformations. With right choice of transformation, a nonstoquastic Hamiltonian with interaction outperform stoquastic similar parameters. We introduce an optimization protocol determine optimal transformation discuss effect suboptimality.